Assessment of Fuzzy Gaussian Naive Bayes for Classification Tasks

摘要

Statistical methods have been used in order to classify data from random samples. In general, if we know the statistical distribution of data, we can use specific classifiers designed for that distribution and expect good results. This work assesses the accuracy of a Fuzzy Gaussian Naive Bayes (FGNB) classifier for tasks using data from five different statistical distributions: Negative Binomial, Logistic, Log-Normal, Weibull and Gamma. The FGNB classifier was recently proposed as a fuzzy extension of Gaussian Naive Bayes for training assessment in virtual environments. Results of assessment are provided and show different accuracy according to the statistical distribution of data.